← Back to Projects Manufacturing

Manufacturing Defect Detection

Computer vision pipeline using ResNet-50 with transfer learning for product defect detection, Grad-CAM localization, active learning, and ONNX Runtime export.

PyTorch
torchvision
OpenCV
ONNX Runtime
FastAPI

Overview

Computer vision pipeline for automated product quality inspection with edge deployment support.

Architecture

  • ResNet-50 with transfer learning for defect classification
  • Grad-CAM for defect localization
  • Active learning with uncertainty sampling
  • ONNX Runtime export for edge deployment
  • FastAPI inference endpoint

Key Features

  • Defect localization with heatmap visualization
  • Active learning to minimize labeling costs
  • Edge-optimized ONNX model export
  • Configurable detection thresholds